The RF PoL: A Special Breed of Point of Load Converters

The air around us all is a sea of radio emissions from a few tens of kHz up to tens of GHz and transmitters' power levels from milliwatt power to megawatts for the biggest ‘longwave’ broadcast stations. We’re all perhaps familiar with Wi-Fi, Bluetooth, Zigbee and many RF-base TV remotes, car keys for example, but an exploding area is 5G, using typically 2.5 – 3.7 GHz bands and extending to millimeter-wave up to 70 GHz. This is driven not only by the surge in personal cell phone users globally, but also by connectivity to the Internet of Things (IoT) in domestic, commercial, and industrial settings. Autonomous vehicles will add to the mix as well, with reports predicting around 160 exabytes of total traffic by 2025 from connected devices, exceeding one million per square kilometer in some locations.

The potential data rates of up to theoretical 10-50 Gb/s promised by 5G are enabled by the higher frequency of operation compared with earlier standards, but this is at the expense of coverage, around 1.5 km at 70 GHz. This means that far more ‘cells’ are needed with all of the associated infrastructures. 5G cells fall into coverage categories, from the highest to lowest power with correspondingly reduced range: ‘metro’, ’micro’, ‘pico’ and ‘Femto’. The metro cell with Multiple Input Multiple Output (MIMO) technology transmits at over 100 W, with the femtocell operating at milliwatt levels.

All Cells Incorporate an RF Power Amplifier

A common feature of all the cell types is an RF power amplifier, the ‘PA’ in Figure 1. This provides power amplification to drive the antenna by taking a small RF signal, for the highest efficiency and radiated signal level. At 4G frequencies and lower, LDMOS transistors have been commonly used and are capable of kW-level output, but for the higher frequency and the lower power requirements of smaller distributed 5G cells, Gallium Nitride (GaN) devices are now preferred for their lower losses, large voltage handling capabilities and excellent thermal performance.

Figure 1: A typical cell RF section outline

In Figure 1, ‘power’ for the PA stage is simplistically shown but is a critical element for the best efficiency and spectral purity of the transmission. An LDMOS PA typically requires 26-32 V and GaN 28-65 V, and the power supply must be capable of responding to fast load changes without significant over- and under-shoot while keeping its nominal voltage accurate. The load varies from transmit to receive and as the transmit power is modulated by data and is dynamically scaled for interference, energy and connectivity management. Even with GaN, RF PA efficiency is only around 60% so for the larger ‘metro’ cells, the PA power supply may need to be able to supply over 200 W or 4 A at 50 V. The power supply will invariably be a switched-mode type and therefore generates its own noise which has to be at a very low level, to avoid creating interference ‘sidebands’ around the wanted transmit signals as a result of PSMR.

With all these constraints, an approach that can be taken is to use a non-isolated Point of Load regulator or RF PoL. PoLs are perhaps more familiar next to CPUs generating sub-1 V rails at tens or hundreds of amps, but the RF PA application has the same requirement for voltage accuracy at the load with fast dynamic response and low noise. An RF PoL, with its high output voltage, however, is a different design from the CPU version. It may be a similar topology, typically a buck converter, but the higher output voltage directly affects efficiency attainable, device ratings and control loop design.  Particularly, noise can be problematic with higher voltage swings and the fast switching edges necessary for high efficiency. 

Evaluating an Example of RF PoL

Suitable RF PoLs are available, such as the ACT43850 from Qorvo. This is a buck controller utilizing wide-bandgap (WBG) transistors with an input range up to 150 V and an output programmable from 20 to 55 V, rated up to 20 A. A feature of the Qorvo device is the high surge current capability, over 20 A, ideal for the peak power demands of an RF PoL. Evaluating the performance of such a part includes checking control loop stability for a given load, which can be done by measuring response times and any over-or under-shoot during transient loads. Zero to 20 A+ is not a realistic load scenario so monitoring output voltage transients in this large-signal loop response situation will give misleading results. A more realistic setup showing the true small-signal response, verifying the control loop stability, would be to apply a higher current and superimpose smaller load steps such as 20 A with 2 A steps, with a customized test arrangement. In practice, because 20 A is a peak rating for this part and cannot be sustained for thermal reasons, steps of 0 A – 20 A – 22 A – 20 A -0 A might be applied with appropriate duty cycles so that temperature rises are constrained. Figure 2 shows an example result using these values on the Qorvo RF PoL.

  Figure 2: Load transient testing of an RF PoL

The output voltage is in yellow, here pulsing 0 V – 50 V – 0 V, and the red plot is the output current. This does show a downward slope from its initial 20 A but that is an effect of the AC-coupled probe. In reality, it is a constant 20 A with a 2 A step upwards in the middle of the plot then a decrease to 20 A then to zero. The output voltage transient from the 2 A step in blue can be seen to be about 20 µs duration with about 150 mV excursion, or just +/-0.3% of Vout, which is a very creditable performance, +/-5% being more typical for low voltage PoLs. The loop response time of this particular RFPoL is so fast that it could even be considered for the implementation of ‘envelope tracking’ via its digital interface. Demonstration of the large-signal response is seen when the load transitions 0 – 20 A and 20 A – 0 A with an excursion that is far from symmetrical. On the positive-going load step, it is underdamped but fast and on the negative step, it is more damped but slow in response to an anomalous ‘kink’ to the waveform. This is an indication that the control loop is operating out of its linear range, perhaps saturating the error amplifier momentarily, then taking time to re-establish correct operating biassing. In the case of the Qorvo part tested, this features synchronous rectification, so the asymmetry is not caused by a change of buck operating modes or discontinuous and continuous conduction. 

Output Noise is Critical in RF PoL Applications

It is always difficult to measure high-frequency noise accurately on switching regulator outputs. Common mode and differential mode elements combine and pickup distorts results. For an RF PA application, what practically matters is the purity of the transmitted signal and the associated levels of ‘spurs’ – extraneous line emissions around the carrier frequency. Measurement of the spectral noise density of the converter under real operating conditions is therefore a better comparison metric, using a spectrum analyzer. An RF PoL operating at full peak power has to be AC coupled to the typically 50-ohm analyzer input. This is done, for example, using a passive high-frequency, high impedance probe that has near-zero capacitance to avoid circuit loading and noise peaking from probe resonance. These probes are often described as ‘PDN’ types as they are effective in measuring noise in the low and ultra-low impedance environments found in power distribution networks.

The probe might typically have x20 attenuation, so needs a following pre-amplifier to raise the signal significantly above the spectrum analyzer noise floor and match into its 50-ohm impedance. Cabling between the probe, preamplifier and analyzer should be verified and included in any calibration routine for the test setup. This can be facilitated with a known precision noise source to ensure that results can be trusted. A suitable calibration and test setup is described in and shown in Figure 3.

Figure 3: Accurate noise measurement test set up for an RF PoL

Typical results are shown in Figure 4 with the lower plot superimposing noise floor from the test setup, demonstrating a good test margin.

Figure 4a/4b: Noise spectral power density plot of a typical RF PoL

Although not related to the plot of Figure 4, the typical effect of noise from an RFPoL on the RF transmission spectrum can be seen in the example of Figure 5, with the PoL switching frequency ‘spurs’ introduced at 500 kHz at about 67 dBm below the carrier with multiples lower still.

Figure 5: A typical cell output spectrum showing ‘spurs’ from RF PoL switching noise.

PoL noise output can be modified with clock frequency ‘dithering’ techniques to spread and reduce the peaks in a noise spectrum - a feature that can be remotely enabled in the Qorvo PoL via an I2C and GUI. PoL clock synchronization with an external signal is also possible to avoid indeterminate ‘beating’ effects. 

Conclusion

DC Power for RF PA stages is critical for best performance. Point of Load converters designed specifically for the application can be a great solution when verified for load transient response and output noise levels. Other potential applications also exist for the RF PoL in medical, measurement, laser power supplies and more.

Click here to learn more about Qorvo and its products on everything RF.

Click here to see DC to DC converters from Qorvo on everything PE.